Sanity check on use of dictionaries

[dopey483]

I am manipulating lots of log files (about 500,000 files and about 30Gb
in total) to get them into a little SQL db. Part of this process is
"normalisation" and creating tables of common data. I am creating
dictionaries for these in a simple {value,key} form.

In terms of memory and performance what are the reasonable limits for a
dictionary with a key and a 16 character string? eg; if I read in one
of my tables from disk into a dictionary, what sizing is comfortable?
100,000 entries? 1,000,000 entries? Lookup times and memory
requirements are my main worries.

(Running Python 2.3.4 on RH Ent, dual-Xeon with 2GB memory)

[Fredrik Lundh]

do******@gmail.com wrote:

I am manipulating lots of log files (about 500,000 files and about 30Gb
in total) to get them into a little SQL db. Part of this process is
"normalisation" and creating tables of common data. I am creating
dictionaries for these in a simple {value,key} form.

In terms of memory and performance what are the reasonable limits for a
dictionary with a key and a 16 character string? eg; if I read in one
of my tables from disk into a dictionary, what sizing is comfortable?
100,000 entries? 1,000,000 entries? Lookup times and memory
requirements are my main worries.

you don't specify what a "key" is, but the following piece of code took
less than a minute to write, ran in roughly two seconds on my machine,
and results in a CPython process that uses about 80 megabytes of memory.

>>d = {}
for i in range(1000000):

.... k = str(i).zfill(16)
.... d[k] = k
....

>>k

'0000000000999999'

since dictionaries use hash tables, the lookup time is usually
independent of the dictionary size. also see:

http://www.effbot.org/pyfaq/how-are-...mplemented.htm

</F>